采用高速数字摄影记录了大气边界层风洞中的沙粒蠕移运动. 通过非负减法、自动阈值法以及双向粒子追踪法的组合, 从图像序列中提取了蠕移运动数字轨迹, 并由此计算得到用于蠕移轨迹类型划分的蠕移运动初速分布. 对两种蠕移轨迹的速度增长率及历时进行了统计分析, 结果表明: 风曳力起动蠕移颗粒运动速度保持性较好, 随着摩阻风速增大, 其运动骤停性及骤停之前的加速性增强; 而碰撞起动蠕移轨迹的速度增长率的分布稳定, 基本不受摩阻风速变化的影响, 且其主要范围为[0,1], 表明碰撞起动蠕移颗粒更易于在运动过程产生速度衰减. 两种轨迹的历时均为高斯分布, 随着摩阻风速增大, 床面运动随机性增强, 轨迹历时趋于减小. High-speed digital photography was adopted to record sand creep motion in atmospheric boundary layer wind tunnel. Algorithms as non-negative subtraction, dynamical threshold and bidirectional particle tracking were applied to abstract digital creeping trajectories out of image sequence. Based on lift-off velocity distribution, statistical analysis of trajectory characteristic parameter including velocity growth rate and duration of trajectory indicated that, majority of flow-initiated particles tended to velocity maintenance or acceleration, and probabilities of both particles’ sudden stagnation and acceleration before stagnation improved as friction velocity U* increased; nevertheless, distribution of velocity growth rate of collision-initiated creeping trajectories behaved stable and much more independent of U* variation, whose concentrating on [0,1] demonstrated the overall velocity-decaying feature of collision-initiated creep motion. Durations of both kinds of trajectory followed the Gauss Distribution, from which a duration-decaying tendency with increase of U* were obtained.